Replicates of soil samples were collected from two sites, Villard and Vénachat, located near Bessines-sur-Gartempe, as detailed in Figure 1 and below in the Materials and Methods section. For each site, radioactive soil samples (U samples) and nearby non-radioactive controls (C samples) were subjected to chemical and mineralogical analyses.

Results for major element chemistry, selected trace elements, pH, and Total Organic Carbon (TOC) analyses are compiled in Table 1. For each series (Vénachat or Villard), TOC content, pH, major elements as well as trace elements (As, Cd, Co, Cr, Cu, Ni, Pb and Zn; not shown) were within the same range between the U and C samples, except for CaO in Vénachat U2-06. As expected, higher uranium concentrations were measured in U samples than in C samples. In Vénachat, C samples exhibited less than 100 ppm of uranium while U samples were 71–87 times more concentrated with values reaching 4,700 ppm. In Villard, the contrast was even greater: the uranium content in C samples, averaging 45 ppm, was 1,000 to 10,000 times lower than in U samples, in which uranium ranged from 14,860 to 255,000 ppm. These high uranium contents are associated with high P₂O₅ concentrations, which is consistent with the mineralogical nature of the soil (see further).

We determined the mineralogy of the soil samples using X-ray Diffraction (XRD). Analyses of Vénachat C-07 and U-07 samples, as well as Villard C-07 and U-07 samples were performed. All soil samples are mostly composed of quartz, orthoclase (K-feldspar), albite (Na-feldspar) and muscovite (data not shown). In Vénachat U-07, a small XRD peak at 8.40 Á is likely related to the (001) reflection of meta-autunite (Ca(UO₂)₂(PO₄)₂ · 4(H₂O)). In Villard U-07, XRD peaks at 8.40, 5.37 and 3.61 Á, indexed as the (001), (101) and (102) planes of meta-autunite could be observed. Additionally, peaks at 9.59, 4.85 and 3.47 Á were interpreted as the (002), (004) and (200) reflections of sabugalite (HAl(UO₂)₄(PO₄)₄ · 16(H₂O)), a common secondary mineral formed in the oxidized zone of uranium veins. These results were confirmed by Scanning Electron Microscopy coupled with Energy Dispersive X-ray Spectroscopy (SEM-EDXS) analyses that detected mineral grains containing P and U on Villard U-09 sample (Figure 2A). We also observed the presence of bright yellow particles in the uranium-rich soil from Villard, a color frequently observed in the autunite group of minerals. In addition to the high P and U contents measured by bulk chemical analyses, these data suggest that a fraction of uranium is under the form of crystalline meta-autunite and associated secondary alteration phases such as sabugalite.

Uranium speciation in the soils was further investigated by X-ray Absorption Spectroscopy (XAS) analysis. Spectra of Vénachat and Villard U-07 soil samples as well as yellow particles extracted from Villard soil were recorded at the U L_III-edge (Figure 3). The position of the absorption edge at ∼17.177 keV is characteristic of the U(VI) oxidation state [31], and the shoulder at 17.188 keV in the XANES region (Figure 3A) is characteristic for the linear uranyl dioxo cation [32]. Thus, the major chemical forms of uranium in both soils and in the yellow extracts are complexes of uranyl cation.

Extracted EXAFS spectra (Figure 3B) of Vénachat and Villard soils display very different features in the region 6–8 Å⁻¹, indicating different uranium speciation. EXAFS spectra of Villard soil and of yellow extracts collected from this soil are quite similar. The best result from shell-by-shell fitting applied to the EXAFS spectrum of these yellow clusters (Figure S1) is given in Table 2. These data are in the range of previously reported values for uranyl-phosphate complexes [32], and may correspond to meta-autunite and/or sabugalite (structural similarities among minerals in the uranyl-phosphate group make EXAFS spectra difficult to distinguish). A contribution at 6.5–7 Å⁻¹ was observed in the EXAFS spectrum in soil from Villard but not in the spectrum of yellow extracts. This suggests the presence of another U-containing phase in this sample, which may be non-crystalline. Attempts to identify this compound by linear combination least-square fitting of the ViU-07 EXAFS spectrum, using uranyl ion, meta-autunite, sabugalite, U-carbonate, and U-acetate as references were unsuccessful.

The EXAFS spectrum of Vénachat U-07 is more complex than that of Villard U-07. Qualitatively, two structures at 5.4 and 7.4 Å⁻¹ signal the presence of a U-phosphate complex, likely the meta-autunite identified by XRD analyses. A second major contribution with beats at 6.6, 8.6 and 9.3 Å⁻¹ (Figure 3B, arrows) could not be fitted properly in the EXAFS spectrum using linear combination least-square fitting. This excludes carbonate-based, silicate, or carboxylate-based species. Since meta-autunite and sabugalite were the only crystalline phases identified using XRD in this soil, the specific features observed in the EXAFS spectrum may come from a non-crystalline uranium-containing compound.

These results suggest that uranium was partly associated with insoluble mineral phases in the soils, raising the question of its bioavailability. To estimate the uranium threshold concentration to which bacterial populations are exposed in situ, the soluble uranium fraction was measured. For Villard and Vénachat samples, approximately 0.1% of the total uranium content was recovered in the soluble fraction after three successive lixiviations with water (Table 1).

Villard U-09 contains the highest uranium grade (25% w/w), and we chose this sample to identify uranium-rich particles and potentially uranium-associated cells. In situ observations by means of SEM-EDXS analysis revealed small particles that contain uranium and phosphate (Figure 2A), as well as carbon-containing objects similar to bacteria in shape and size (Figure 2B), likely corresponding to individual or filamentous bacterial cells. Several cell-shaped objects were entirely covered by particles identified as uranyl-phosphate precipitates by EDXS (Figure 2C).

Taken together, the analyses of the Villard and Vénachat soil samples indicate that the uranium-rich and control samples differ essentially in their contrasted uranium content, which is linked to the presence of mineral phases such as meta-autunite and sabugalite, and that bacteria are exposed to uranium in situ. Thus, comparing the natural environment of these two sites is ideal for estimating the impact of uranium on bacterial communities by a comparative study.

Soil samples were collected from the Villard and Vénachat sites, located near Bessines-sur-Gartempe, Limousin, France. This region is well-known for the presence of important natural uranium deposits.

Vénachat is located 15 km south of Bessines-sur-Gartempe. A uranium-rich area (approximately 2 m×0.5 m) was detected with a Geiger counter, 10 m from the roadside, at the foot of a granitic outcrop exhibiting a uraniferous vein. In October 2006, 8 independent samples (approximately 50 g each) were collected on the soil surface at 2 different positions in this area (U1 and U2). A series of 8 non-radioactive control samples (C) were collected in close vicinity (<10 m apart; approximately 50 g each). Each sample was homogenized by manual mixing before DNA extraction. Subsequently, U1, U2 and C samples were pooled and mixed before pH measurement, bacteria cultivation and a partial chemical analysis. In October 2007, 8 radioactive and 8 control samples (approximately 50 g each) were collected randomly from the same areas as 2006. One DNA extraction was performed for each radioactive soil sample (1 g per extraction), after which all DNA were pooled (designated as sample U-07). DNA was similarly extracted and pooled from control samples (designated as sample C-07). Soil samples were then pooled to give U-07 and C-07 samples. Subsequent bacterial cultivation, DNA extraction and in-depth mineral and chemical analyses were performed on these samples.

At Villard, 2 km northeast of Bessines-sur-Gartempe, three separate soil samplings were performed in October 2006, October 2007 and April 2009. A very restricted radioactive spot (approximately 30 cm in diameter) was detected in a ditch along a field. Different soil samples (approximately 100 g) were collected from this area: U-06 (2006), U-07 (2007) and U-09 (2009). Samples were collected at 10–20 cm depth and stored in sterile tubes. The presence of bright yellow particles was observed in these samples. Non-radioactive control samples C-06 (2006), C-07 (2007), and C1-09; C2-09; C3-09; C4-09; C5-09 (2009) were collected in close vicinity (50 cm to 5 m apart) within the ditch. Each sample was homogenized by manual mixing and pH measurements, and DNA extraction and bacteria cultivation were processed immediately after collection. Samples were then stored at 4°C until chemical analyses could be performed.

To estimate the number of culturable aerobic bacteria, soil sample suspensions were spread on 0.1× TSB agar plates (TSB, Difco Laboratories). The number of CFU per g of soil was measured after incubation at 30°C for two weeks.

For pH determination, 5 g of soil were mixed with 5 ml of sterile deionized water, shaken for 1 h and decanted. pH of the supernatant was measured. Elemental composition of soil samples was determined after homogenization and mineralisation by ICP-AES for major elements (NFX31-147/NF EN ISO 11885 method) and ICP-MS for traces (NFX31-147/ISO 17294-2 method). The uranium content was measured on 0.5 g of soil samples. The Total Organic Carbon (TOC) concentration was assayed by specific equipment (Leco system - NF ISO 10694-2 method).

In order to determine the amount of water-extractable uranium, soil elutriates were prepared according to the AFNOR [46] standard protocol: 2 g of dried soil (24 h at 105°C) were mixed with 20 ml of deionised water, shaken 16 h at 20°C, centrifuged at 3000 g for 15 min and the supernatant was collected; this procedure was repeated three times. Samples for chemical analyses were acidified with HNO₃ and stored at 4°C in the dark until analysis. Uranium content was measured in each elutriate sample by ICP-MS.

Soil samples VeC-07, VeU-07, ViC-07 and ViU-07 were finely ground using an agate mortar and pestle. X-ray diffraction (XRD) patterns were recorded with a Panalytical X'Pert Pro MPD® diffractometer mounted in the Debye–Scherrer configuration, using an elliptical mirror to obtain a high flux, parallel incident beam and an X'Celerator® detector to collect diffracted beams. Data were recorded with a monochromatic CoKαbeam (λ = 0.17889 nm) in continuous scan mode within a (3°–120°) 2θ range with steps of 0.0167° and a counting time of 260 s per step.

XAS experiments were performed at the U LIII-edge (17.166 keV) on the BM30B beamline [47] of the European Synchrotron Radiation Facility, Grenoble, France. Experimental details are given in Text S1. Spectra were normalized and EXAFS oscillations were extracted using the Athena code [48]. The resulting EXAFS curves were weighted by k3 and qualitatively analyzed by comparison between spectra of samples and references. Quantitative analysis was performed in one case using simulations performed with Artemis code.

One gram of the soil sample ViU-09 was incubated 2 days at 30°C in 1 ml 0.1 X TSB and the resulting suspension was deposited on a lacey 300-mesh copper grid without coating. SEM observations were performed on a Zeiss Ultra 55 FEG-. The microscope was operated at 15 kV at a working distance of 7.1 mm. Images were acquired in secondary electron mode using an Everhart Thornley detector. Energy dispersive x-ray spectrometry (EDXS) analyses were performed using an EDS QUANTAX microanalyzer with Esprit, Hypermap software that allowed acquisition of x-ray maps and drift correction.

For Vénachat soil samples, DNA was extracted from 1 g aliquots using a PowerSoil™ DNA Isolation Kit (MO Bio, USA). The yield was approximately 500 ng of DNA per gram of soil. In 2006, 8 independent DNA extractions were performed for 3 series of soil samples: U1-06, U2-06 (radioactive) and C-06 (control). For each series, 5 DNA samples were pooled (samples “p”) while 3 remained independent (samples “a, b and c”) for a subsequent DGGE analysis. In 2007, 8 DNA extractions from each radioactive (U-07) and control (C-07) soil were performed and pooled before DGGE analysis.

For Villard samples, no DNA could be recovered from the radioactive samples with the PowerSoil™ kit. An alternative SDS-based extraction method was used as described by Zhou et al. [49], starting with two 10 g aliquots of soil in 2006, and 20 g in 2007 and 2009. A final DNA purification step was added using the DNeasy tissue kit (Qiagen), following the manufacturer's instructions. The yield was approximately 30 ng of DNA per gram of soil for radioactive samples and 400 ng of DNA per gram of soil for controls.

For the DGGE analysis, metagenomic DNA extracted from soil samples was used as a template for PCR amplification of 16S rRNA genes using the universal primers set fD1 and S17 (Table S1). The 1500bp fragments were then re-amplified using primers that target Bacteria, Alphaproteobacteria, Betaproteobacteria, Gammaproteobacteria, Actinobacteria, Firmicutes and CFB to generate DGGE-compatible fragments. Experimental conditions for PCR reactions and DGGE electrophoresis are detailed in Text S2.

DGGE data obtained from the Genetools software were converted into a table summarizing the band presence and intensity. The total intensity of bands from each lane was normalized to 100. Principal Component Analysis (PCA) was then performed with the R-software. Statistical ellipses representing 90% confidence on PCA plots were used to compare DGGE profiles.

For Villard samples, DGGE bands characterizing uraniferous profiles were excised from the gel using sterile scalpel blades. DNA was eluted in sterile water at 4°C overnight then re-amplified with primers P1 and COM2. Direct sequencing of the PCR products revealed the presence of overlapping DGGE bands with heterogeneous sequences, a feature that has been reported frequently in DGGE studies [50]-[53]. The PCR products were then cloned using the TOPO PCR cloning kit (Invitrogen Life Technologies, Carlsbad, CA, USA) following the manufacturer's instructions, prior to their sequencing. Finally, a phylogenetic analysis was performed on all 144 clone sequences obtained.

In order to assign taxonomy, clone sequences were compared to the Silva 106 reference sequences (http://www.arb-silva.de/). Each clone sequence was globally aligned (Needleman-Wunsch algorithm) to each reference sequence and a percentage of similarity was calculated using a penalty of 1 for mismatch, open a gap and extend a gap. This approach is more rigorous and precise than using BLAST, which does local alignments and searches only for exact words of 7. First we looked for hits with ≥99% similarity, in which case we calculated a consensus for their taxonomy. In the event that no hit was found with ≥99% similarity, we successively lowered the threshold in a step-wise fashion to determine at what level we could assign a taxonomy. This process was repeated until an 80% threshold was reached. Thus, the range of similarity percentages we successively looked for were: 99, 98.5, 98, 97.5, 97, 96, 95, 94, 93, 92, 91, 90, 85 and 80. In a final step, a consensus taxonomy was calculated between the hits retained from the above process. For example, a clone sequence could be assigned to a defined genus if all hits share the same genus, or simply to "Bacteria" if the selected hits conflicted at the phylum level. The sequences for each best hit from Silva were retrieved and aligned with the clone sequences using Muscle [54]. A first tree was built using conserved domains, and sequences were reordered as they occur in this tree using SeaView [55]. Alignments were checked and manually adjusted when required. Domains common to all sequences, and which show no evident sign of homoplasy, were selected and a final tree was constructed using a Kimura two parameters correction and BioNJ as implemented in SeaView.

In order to find similar sequences from polluted environments, the same similarity searches were performed with the EMBL database (release 107). In this case, every hit with greater than 95% similarity was considered valid. Each entry was analyzed for the presence of a series of keywords (uranium, radionuclide and iron). Finally the tree was displayed using [56] and information retrieved as described above was used to build Figure S2.

The 16S rRNA gene sequences determined in this study were deposited in GenBank under Accession No. JN032139–JN032282.